dive into mark

PyBlogger 0.7 is out. It has a new function, listPosts(), which gets information (date created, user ID, post ID, and text) of recent posts. It also has an object model, Blog.py:

"""Object model for Blogger

This is a high-level object-oriented interface for blogger.py,
which defines all the available functions exposed by Blogger.com's
XML-RPC API.

Example:
>>> user = Blog.User("YOUR_BLOGGER_USERNAME", "YOUR_BLOGGER_PASSWORD")
>>> blogs = user.blogs                # list of all blogs
>>> for blog in blogs:
...     print "Blog ID:", blog.id     # internal Blogger.com blog ID
...     print "Blog name:", blog.name # title of blog
...     print "Blog URL:", blog.url   # base URL of blog

>>> blog = blogs[0]                   # get reference to user's first blog
>>> posts = blog.posts                # list of most recent posts (up to 20)
>>> for post in blog.posts:
...     print "Post ID:", post.postid        # internal Blogger.com post ID
...     print "Post date:", post.dateCreated # date created, in tuple format
...     print "Posted by:", post.userid      # internal Blogger.com user ID
...     print "Post text:", post.content     # text of post

>>> posts.append("Ping.")             # post new entry to blog
>>> len(posts)                        # count posts
>>> posts[-1].content = "Pong."       # edit text of most recent post
>>> del posts[-1]                     # delete most recent post

>>> html = blog.template.main         # get HTML of main blog entry template
>>> blog.template.main = html         # set HTML template for main blog entries
>>> html = blog.template.archiveIndex # get HTML of archive index template
>>> blog.template.archiveIndex = html # set HTML template for archive index
"""

I’m already working on a new version which will have better unit tests that can be run by anybody. The current unit tests work by actually calling the real Blogger.com server and then using urllib to check the results. There are several problems with this:

• The calls are asynchronous, so I have to sit in a wait loop waiting for new posts to show up or edited posts to take effect
• It ties up Blogger.com unnecessarily (Evan has enough problems without me running 31 unit tests over and over)
• It requires my username and password, which I don’t give out (I store them in a separate file and read them in), so only I can run the unit tests.
• It’s inflexible. You couldn’t just come along and change a few numbers and put in your own username and password and expect the unit tests to pass; the test cases are very much tied into my blogs, my templates, and so forth.

To solve these problems, I decided to start over from scratch and make my own fake XML-RPC server. I figured out how to override the crucial part of xmlrpclib to redirect it to my own transport class. It’s hard to explain (wow, I don’t usually say that), but the upshot of it is that blogger.py thinks that it’s communicating with the Blogger.com server, but in reality it’s communicating with a locally-defined class (part of the test suite) that acts like the Blogger.com server.

This has several advantages:

• I don’t have to change the code I’m testing in order to test it.
• It’s entirely self-contained and relies on no real usernames or passwords, so anyone can run it.
• It doesn’t tie up any Blogger.com resources, so Evan will be happier.
• It’s a real unit test, not a functional test; it tests only that the blogger.py functions pass the right XML to the server and handle the response correctly. The old “live” unit tests would fail occasionally because of errors on the Blogger server. The new unit tests isolate the code they’re testing.

The downside, of course, is added complexity. I kept the fake server class as simple as possible, but it’s still extra code that could have its own set of bugs. If a unit test fails, is it because blogger.py is wrong or because the server class isn’t correctly serving up fake data? On the whole, though, I think the benefits are worth the added complexity. Nobody said good programming was easy.

§

Respond privately

I am no longer accepting public comments on this post, but you can use this form to contact me privately. (Your message will not be published.)

§